SETI and a old server

[Desslok]

I am running SETI and I have an old dual XEON PIII workstation, my question is which Xeon proc would be better suited to crunch the numbers? One with a slower clock speed and a large cache or one with a faster clock speed and a smaller cache?

 

[Sukhoi]

Distributed computing forum is where you want this question. 🙂 Generally you'll want 512 KB cache so the whole WU fits in there, and 1 MB might be even better. But what kind of a speed range are we talking about?

 

[Desslok]

DOE sorry about that.

They are 2x 667 with a 256 cache and 2x PII 500 with a 2meg cache

Mods please move if needed.

 

[ColinP]

Tricky one to call, that extra Mhz and the 133 FSB will really help on one side, but that 2Mb cache will really help on the other...

My guess is about the same with the 667 maybe winning slightly,

a single 667 will do a WU in between 8 & 9 hours, dual - maybe 2 x WU in 10 to 11 hours ?

You'll have to run it one both systems or we'll never know !!!

let us know

🙂

Col

 

[networkman]

Based on experience here, I'd say the 133fsb will more than compensate for the smaller cache size of the 667 chips.

I have a dual p2-400 Xeon rig here - each CPU having 2meg of L2 cache, and I'd much rather have a faster FSB than the 100 these are at; also, cool as it sounds, 2meg of cache has no discernible performance improvement over 1meg of cache. 😱

 

[BadThad]

I'm with NWM, basically RAM speed plays a HUGE part...the faster the RAM, the faster SETI completes.

 

[Desslok]

Ok thanks for the info guys. I will run tests using both sets of procs and let you know.

 

[Sukhoi]

Was the Xeon cache always full-speed?

 

[BadThad]

Sukhoi - Yes, that was the main thing about them before integrating the cache into the die. With the 3.06 GHz out, I'm not sure XEON has any advantage today sans SMP ability.

 

[Desslok]

Originally posted by: badthad
Sukhoi - Yes, that was the main thing about them before integrating the cache into the die. With the 3.06 GHz out, I'm not sure XEON has any advantage today sans SMP ability.

Does it even have the SMP advantage now with Hyper Threading out? I know that the software companies have to enable it for it to work as advertise, but after they do I wonder what Intel will do. If anything

 

[Sukhoi]

Ah, thanks badthad.

I actually think the performance of both processors will be quite close then. I'm interested to see Desslok's results.